Giving K Pro a full-text evidence layer
Lucas Brechot, Senior Product Manager, AI Products, Owkin
Owkin, a biotechnology company building AI tools for biopharma drug discovery, integrated the Consensus MCP into K Pro, its Agentic AI Scientist. The upgrade swapped a curated collection of PubMed abstracts for 200M+ full-text articles updated weekly, with hallucination safeguards built in, and shipped to users in days.
1
the company
An Agentic AI Scientist for biopharma teams
Owkin is a biotechnology company building AI tools that accelerate drug discovery and development for biopharma. Its flagship product, K Pro, is an Agentic AI Scientist that orchestrates specialized agents across clinical trial design, patient and population decisions, and early portfolio choices. This is work that traditionally requires weeks of literature review and cross-functional input from research scientists, biostatisticians, and clinical experts.
2
the problem
An AI scientist needs more than abstracts
The challenge for any agentic system operating in biomedicine is grounding. An AI scientist is only as good as the evidence base it can reason over.
Early versions of K Pro relied on a curated collection of PubMed abstracts, a serviceable foundation, but one that left agents reasoning from titles and summaries rather than full-text findings, with limited filtering for study quality and no built-in defense against hallucination.
3
The solution
Consensus became a research-grade evidence layer inside K Pro
Owkin integrated Consensus through MCP, giving K Pro access to a much broader and fresher literature base without rebuilding that infrastructure internally.
“The Consensus MCP slotted directly into our agentic stack and gave K Pro’s capabilities a dramatic upgrade overnight,” says Lucas Brechot, Senior Product Manager of AI products at Owkin. “We went from maintaining a collection of PubMed abstracts to 200M+ full-text articles updated weekly, with hallucination safeguards built in. It’s exactly the kind of best-in-class building block we want to compose with, rather than rebuild.”
4
in production
The integration fit the way K Pro already exposes tools
K Pro accesses Consensus through its MCP, exposing it to the orchestrator the same way Owkin exposes its other tools. That architectural fit mattered. The integration was plug-and-play, and the team shipped the upgrade to users in days.
Agents now ground their reasoning in full-text literature with programmatic controls for recency, citation count, and journal reputation, the kind of filters a human researcher applies instinctively when evaluating a paper’s relevance and credibility.
For Owkin’s customers (biopharma teams making high-stakes decisions about which trials to run, which patient populations to target, and which portfolio bets to place), the upgrade compounds. Better evidence retrieval means better agent reasoning, which means K Pro’s recommendations carry more weight in the rooms where multimillion-dollar development decisions actually happen.
5
The bigger picture
Owkin treated evidence retrieval as infrastructure to compose with
Building an agentic AI scientist is hard enough without also building a research-grade evidence layer from scratch. By treating Consensus as infrastructure rather than something to replicate internally, Owkin’s team stays focused on the orchestration, agent design, and domain expertise that make K Pro distinctive.
